This article presents a comparative study on different types of robust design optimization methods for electrical machines. Three robust design approaches, Taguchi parameter design, worst-case design and design for six-sigma, are compared for low-dimensional and high-dimensional design optimization scenarios, respectively. For the high-dimensional scenario, the computational burden is normally massive due to the robustness evaluation of a huge number of design candidates. To attempt this challenge, as the second aim of this paper, a space reduction optimization (SRO) strategy is proposed for these robust design approaches, yielding three new robust optimization methods. To illustrate and compare the performance of different robust design optimization methods, a permanent magnet motor with soft magnetic composite cores is investigated with the consideration of material diversities and manufacturing tolerances. 3-D finite element model and thermal network model are employed in the optimization process and the accuracy of both models has been verified by experimental results. Based on the theoretical analysis and optimization results, a detailed comparison is provided for all investigated and proposed robust design optimization methods in terms of different aspects. It shows that the proposed SRO strategy can greatly improve the design optimization effectiveness and efficiency of those three conventional robust design methods.